The role and functions of miRNAmediated circuits in the human
regulatory network
Anacapri 29/5/2012
Michele Caselle – University of Torino and INFN
caselle@to.infn.it
Plan of the talk
1. Introduction: Gene Regulation and Regulatory Networks
2. Bioinformatic analysis:
construction of mixed Feed Forward Loops (FFL)
3. FFL Assessment: Network motifs and GO analysis
4. FFL Function: Stochastic equations and noise damping effect
5. Examples: c-Myc, PTEN…
Work in progress:
6. MiRNA mediated self-loops
7. Sponge loops
References
• D. Cora’, A. Re, D. Taverna and M. Caselle
“Genome-Wide Survey of MicroRna-Transcription Factor
Feed-Forward Regulatory Circuits in Human”
Molecular BioSystems. 2009 Aug; 5(8):854-67.
• M.Osella, C. Bosia, D. Cora’ and M. Caselle
“The role of incoherent microRNA-mediated FFL in noise buffering
PloS Computational Biology (2011) 7(3): e1001101
• M. El Baroudi, D. Cora’, M.Osella, C. Bosia, and M. Caselle
“A curated database of miRNA mediated Feed Forward
Loops involving MYC as Master Regulator”
PloS ONE (2011) 6(3):e14742
• F. Eduati, B. Di Camillo, M.Karbiener, M Scheideler, D. Cora’,
M. Caselle, G. Toffolo
“Dynamic modeling of miRNA-mediated FFLs”
J Comput. Biol. 2012 Feb;19(2):188-99
Gene Regulation
Gene expression is tightly regulated. All cells in the body carry the full set of
genes, but only express about 20% of them at any particular time. Different
proteins are expressed in different cells (neurons, muscle cells....) according to
the different functions of the cell.
Among the various regulatory
steps the most important ones
are:
 transcriptional control,
by Transcription Factors.
 post-transcriptional control,
by microRNAs.
Alberts, Molecular Biology of the Cell
Transcription Factors
and miRNAs
• Regulation of gene expression mainly mediated by:
Transcription Factors (TFs): proteins
binding to specific recognition motifs
(TFBSs) usually short (5-10 bp) and
located upstream of the coding region
of the regulated gene.
Wassermann, Nat. Rev. Genetics
MicroRNAs (miRNAs) are a family of
small RNAs (typically 21 - 25 nucleotide
long) that negatively regulate gene
expression at the posttranscriptional
level, (usually) thanks to the “seed”
region in 3’-UTR regions.
MicroRNAs as regulatory
genes
MiRNAs expression is regulated by the same TF which regulate all the
other genes
Regulation by miRNAs is a combinatorial process. Each miRNA is
expected to control from one to hundreds of targets while a given mRNA
can be under control of many different miRNAs. Usually miRNA binding
sites are overrepresented in the 3’-utr sequence of target genes.
Transcription Factors and miRNAs share very similar regulatory strategies.
The main difference between the two is that while TF act as a sort of on/off
switch, it seems that the miRNA role is to fine tune the gene expression.
Regulatory Networks 1
Key 1 --> TFs are themselves proteins produced by other genes, and they act
in a combinatorial way, resulting in a complex network of interactions
between genes and their products.
--> Transcriptional Network
miRNAs also act in a combinatorial and one-to-many way, and,
moreover, are transcribed from same POL-II promotes of TFs.
--> Post-Transcriptional Network
miRNA X
Gene E
Gene F
QuickTime™ e un
decompressore TIFF (Non compresso)
sono necessari per visualizzare q uest'immag ine.
Protein E
Regulatory Networks 2
Key 2 -->
Biological functions are performed by groups of genes which act in
an interdependent and synergic way. A complex network can be
divided into simpler, distinct regulatory patterns called network
motifs, typically composed by 3 or 4 interacting components which
are able to perform elementary signal processing functions.
TF
miRNA
target gene
....
MiRNA mediated FFLs
Several methods exist to study, separately TF-related and
microRNA-related regulatory networks, but comparable information
is lacking to explicitly connect them.
The main goal of our project is to infer and then combine the two
networks looking in particular for Mixed Feed-Forward Regulatory
Loops --> a network motif in which a master Transcription Factor
(TF) regulates a miRNA and together with it a set of Joint Target
coding genes.
TF
QuickTime™ e un
decompressore TIFF (Non compresso)
sono necessari per visualizzare quest'immagine.
Joint
Target
Hornstein E, Shomron N, Nat Genet 38 Suppl:S20–4 (2006).
miR
Pipeline
human core promoters
human 3’-UTR exons
non-redundant set of
human core promoters
-900 / +100 around TSS
non-redundant set of
full length 3’-UTRs
(protein-coding genes)
(protein-coding + miRNA genes)
Oligo analysis
Oligo analysis
sets of human genes
mouse
promoters
sets of human genes
mouse
conserved
overrepresentation
conserved
overrepresentation
3’-UTRs
regulatory oligos in human promoters and 3’-UTRs
relevance
to cancer
Mixed Feed-Forward regulatory Loops
external annotations
Gene
Ontology
Results
Human Transcriptional Network --> Fixing 0.1 as FDR level, we obtained a catalogue of
2031 oligos that can be associated to known TFBSs for
a total of 115 different TFs.
--> target a total of 21159 genes
(20972 protein-coding and 187 miRNAs)
Human Post-Transcriptional Network --> Fixing 0.1 as FDR level, we obtained a
catalogue of 3989 oligos (7-mers). 182 of them turned out
to match with at least one seed present in 140 mature
miRNAs.
--> target a total of 17266 genes
Human mixed FFLs catalogue --> We were able to obtain a list of 5030 different “single
target circuits”, corresponding to 638 “merged circuits”.
TF
JT 1
--> involving a tTotal of 2625 joint target genes (JTs),
101 TFs and 133 miRNAs.
miR
JT 2
# of JTs ranged from 1 to 38.
JT …
Circuits assessment 1
functional analysis
We analyzed each one of the 638 merged circuits looking for an enrichment in Gene
Ontology categories in the set of their joint targets.
To assess this enrichment we used the standard exact Fisher test with a p-value threshold
p < 10−4.
we end with a list of 32 merged mixed Feed-Forward Loops (corresponding to 380
single-target FFLs). These circuits involve a total of 344 JT protein-coding genes,
24 TFs and 25 mature miRNAs.
QuickTime™ e un
decompressore TIFF (Non compresso)
sono necessari per visualizzare q uest'immagine.
Circuits assessment 1
functional analysis
QuickTime™ e un
decompressore TIFF (Non compresso)
sono necessari per visualizzare quest'immagine.
QuickTime™ e un
decompressore TIFF (Non compresso)
sono necessari per visualizzare q uest'immagi ne.
QuickTime™ e un
decompressore TIFF (Non compresso)
sono necessari per visualizzare quest'immagine.
--> various aspects of organism differentiation and development
Circuits assessment 2: looking for
cancer related genes
In these last few years it is becoming increasingly clear that miRNAs play
a central role in cancer development (e.g. Blattener Mol Syst. Biol. 2008).
We filtered our results looking for
FFLs containing at least two cancer
related miRNA or target gene.
Sources: oncomiRs reported in
-
Esquela-Kerscher and FJ Slack,
Nat Rev Cancer 2006
- Zhang et al, Dev Biol, 2007
cancer genes reported in
- Cancer Gene Census database.
QuickTime™ e un
decompressore TIFF (Non compresso)
sono necessari per visualizzare q uest'immag ine.
Circuits assessment 3: mixed
FFLs as network motifs
Elementary regulatory circuits (the so called ”network motifs”)
were shown
to be over-represented in transcriptional networks.
(Milo et a., Science 2002, Shen-Orr et., Nat Genetics 2002)
In order to quantify the overrepresentation of our mixed FFLs
we perfomed various randomization tests.
- Complete
node replacement, Z = 9.2
- Random reshuffling of miRNA promoters and seeds, Z = 8.1
- Edge Switching, Z = 8.4
FFLs are over-represented
Functional role of mixed FFLs
Depending on the type of transcriptional regulation
(excitatory or inhibitory) exerted by the master TF
on the miRNA and on the targets, FFLs may be
classified as
• incoherent (“type I” FFLs), or
• coherent (“type II” FFLs).
Type I and II FFLs
Possible biological role for mixed TF/miRNA network motifs:
TF
TF
Joint
Target
miR
Joint
Target
TF
Joint
Target
miR
TF
miR
type I circuits
Joint
Target
miR
type II circuits
•
Type II (coherent) circuits lead to a reinforcement
of the transcriptional regulation at the posttranscriptional level and might be important to
eliminate the already transcribed mRNAs when the
transcription of a target gene is switched off.
• Type I (incoherent) circuits allow for a fine tuning
of gene expression, setting the optimal functional
value of the protein through a miRNA repression
Additional role: noise damping
Fine tuning is useless without a tight control of
cell to cell fluctuations.
Type I (incoherent) FFLs can also stabilize the
steady state production of the target protein by
damping translational and transcriptional
fluctuations.
In a simple TF-target interaction any fluctuation of
master TF could induce a non-linear increase in the
amount of its target products. The presence, among
the targets, of a miRNA which downregulates the other
targets might represent a simple and effective way to
control these fluctuations.
Study of protein fluctuations via
stochastic equations
The only way to address this issue is to describe
the FFLs in terms of stochastic equations and to
compare the results with those obtained with
that of a standard transcription + translation
process
In both cases fluctuations are proportional to the
mean number of proteins produced by a single
mRNA. This number is a function of the miRNAmRNA affinity.
Stochastic equations for gene
expression: two steps model.
(Shaharezaei V, Swain PS PNAS (2008) 105, 17256)
This model assumes that the promoter is always active and
so has only two stochastic variables: the number of mRNAs
and the number of proteins
The probability of having m mRNAs and n proteins
at time t satisfies the master equation:
The master equation can be rewritten as a differential equation
using the generating function:
Setting:
and
we find:
If we assume that the protein lifetime is much
longer than that of the mRNA then the equation
simplifies (the mRNA is at steady state) and can be
solved exactly:
leading to an exact expression for the probability
distribution:
which at steady state becomes the well known
negative binomial distribution:
The corresponding mean value and fluctuations of the
number of proteins are:
Where b is the mean number of proteins produced by a
single mRNA (burst parameter). Fluctuations only
depend on the burst parameter b.
The same analysis can be performed in the case of the
inchoerent FFL, leading to a relevant reduction of noise
The noise reduction can be traced back to two parallel
mechanisms:
- The different efficiency of the mRNA translation in the two
cases:
noise reduction is a function of the miRNA-mRNA affinity
-The correlated fluctuations of miRNA and target under
fluctuations in the transcriptional efficiency of the master
Transcription Factor
Master equation for the incoherent FFL
TF
miR
Joint
Target
5
Master Equation
• The first two moments
can be calculated with
the moment generating
function method.
• Non linear model
6
Noise Buffering I
8
Noise Buffering II
10
Optimal noise reduction for intermediate values of
miRNA/mRNA affinity
FFLs with C-Myc as master TF
We developed a parallel dataset of mixed FFLs, with C-Myc as master
regulator in which each regulatory interaction is (independently)
experimentally validated
TF
Joint
Target
TF
miR
miR
Joint
Target
Zeller et al. Database (http://www.myc-cancer-gene.org/)
TransmiR Database (http://202.38.126.151/hmdd/mirna/tf)
Diana Tarbase (http://diana.cslab.ece.ntua.gr/tarbase/)
Enrichment test with experimentally
validated interactions
Example : regulation of PTEN
Example 2: regulation of VEGF
CircuitsDB
We have developed a web-based graphical interface to get free
access to the database of mixed FFLs.
MiRNA mediated Selfloops
The same analysis can be extended to the simplest
possible mixed circuit: the miRNA mediated selfloop in
which an intronic miRNA regulates its host gene.
Also in this case we find fine tuning and noise reduction
properties. Moreover this circuit, depending on the
values of the parameters, is able to perform:
acceleration of response times,
stabilization of the switch on and off dynamics,
fold-change detection
Comparison of noise buffering properties of simple regulation,
transcriptional selfloop and miRNA mediated selfloop
“Sponge-like” interactions
RNA transcripts can cross-talk by competing for common
microRNAs. These transcripts act as “sponges” for the
miRNAs thus inducing an indirect regulatory interaction on
their partners
Sumazin et al. Cell 2011
Cesana et al. Cell 2011
Tay et al. Cell 2011
Example of a sponge like interaction
(from Cesana et al Cell 2011)
“Sponge” loops
miR
TF
Joint
Target
Two main functions:
- Enhance and speed up target protein production
- Correlate target-TF fluctuations: homeostatic effect
Conclusions
The main purpose of this work was to:
--> systematically investigate connections between transcriptional and
post-transcriptional network interactions, in the human genome.
--> we designed a bioinformatic pipeline, mainly based on sequence analysis
of human and mouse genomes, to obtain (and validate) a list
of mixed Feed-Forward Loops and mixed selfloops.

we have shown, solving the stochastic equation which describes
these FFL that the effect of the interfering miRNA is to damp
the intrinsic noise in protein production.
The main outcomes of this work are available in a public database:
http://biocluster.di.unito.it/circuits/
Collaborators
A. Riba, M. El Baroudi
Dep. of Theoretical Physics
University of Torino
M. Osella
P. & M. Curie Univ. Paris
C. Bosia .
HUGEF
Torino
D. Corà
IRCC
Candiolo (TO)
A. Re
CIBIO
University of Trento
D. Taverna
Dep. of Genetics, Biology and
Biochemistry and M.B.C.
University of Torino